47 results about "Adaptive resonance theory" patented technology

A fault detection system for use with a solar hot water system may include a data acquisition module which may, in turn, include a plurality of sensors. Input data may include a sensed condition. The system may also include a neural network to receive the input data which may be a multi-layer hierarchical adaptive resonance theory (ART) neural network. The neural network may perform an analysis on the input data to determine existence of a fault or a condition indicative of a potential fault. The fault and the condition indicative of the potential fault are prioritized according to the analysis performed by the neural network. A warning output relating to the fault and the condition indicative of the potential fault is generated responsive to the analysis, and is displayed on the user interface.

Techniques are disclosed for identifying anomaly object types during classification of foreground objects extracted from image data. A self-organizing map and adaptive resonance theory (SOM-ART) network is used to discover object type clusters and classify objects depicted in the image data based on pixel-level micro-features that are extracted from the image data. Importantly, the discovery of the object type clusters is unsupervised, i.e., performed independent of any training data that defines particular objects, allowing a behavior-recognition system to forgo a training phase and for object classification to proceed without being constrained by specific object definitions. The SOM-ART network is adaptive and able to learn while discovering the object type clusters and classifying objects and identifying anomaly object types.

Techniques are disclosed for analyzing a scene depicted in an input stream of video frames captured by a video camera. In one embodiment, e.g., a machine learning engine may include statistical engines for generating topological feature maps based on observations and a detection module for detecting feature anomalies. The statistical engines may include adaptive resonance theory (ART) networks which cluster observed position-feature characteristics. The statistical engines may further reinforce, decay, merge, and remove clusters. The detection module may calculate a rareness value relative to recurring observations and data in the ART networks. Further, the sensitivity of detection may be adjusted according to the relative importance of recently observed anomalies.

A bio-inspired actionable intelligence method and system is disclosed. The actionable intelligence method comprises recognizing entities in an imagery signal, detecting and classifying anomalous entities, and learning new hierarchal relationships between different classes of entities. A knowledge database is updated after each new learning experience to aid in future searches and classification. The method can accommodate incremental learning via Adaptive Resonance Theory (ART).

A sequence layer in a machine-learning engine configured to learn from the observations of a computer vision engine. In one embodiment, the machine-learning engine uses the voting experts to segment adaptive resonance theory (ART) network label sequences for different objects observed in a scene. The sequence layer may be configured to observe the ART label sequences and incrementally build, update, and trim, and reorganize an ngram trie for those label sequences. The sequence layer computes the entropies for the nodes in the ngram trie and determines a sliding window length and vote count parameters. Once determined, the sequence layer may segment newly observed sequences to estimate the primitive events observed in the scene as well as issue alerts for inter-sequence and intra-sequence anomalies.

Techniques are disclosed for detecting the occurrence of unusual events in a sequence of video frames Importantly, what is determined as unusual need not be defined in advance, but can be determined over time by observing a stream of primitive events and a stream of context events. A mapper component may be configured to parse the event streams and supply input data sets to multiple adaptive resonance theory (ART) networks. Each individual ART network may generate clusters from the set of inputs data supplied to that ART network. Each cluster represents an observed statistical distribution of a particular thing or event being observed that ART network.

A sequence layer in a machine-learning engine configured to learn from the observations of a computer vision engine. In one embodiment, the machine-learning engine uses the voting experts to segment adaptive resonance theory (ART) network label sequences for different objects observed in a scene. The sequence layer may be configured to observe the ART label sequences and incrementally build, update, and trim, and reorganize an ngram trie for those label sequences. The sequence layer computes the entropies for the nodes in the ngram trie and determines a sliding window length and vote count parameters. Once determined, the sequence layer may segment newly observed sequences to estimate the primitive events observed in the scene as well as issue alerts for inter-sequence and intra-sequence anomalies.

The invention is a method for detecting out-of-specification quality of a product and estimating actually measured value of the product. The method for detecting the out-of-specification quality of the product is to construct a fault detection (FD) model representing the relationship between technological parameters and the out-of-specification of the product by applying a classification and regression tree (CART) method so as to detect the condition that the process data parameters are all in the specification (normal) but the product quality goes beyond the specification on a production line in real time. The method for estimating the actually measured value of the product is to establish a measured data quality index (DQIy) model according to an adaptive resonance theory 2 (ART2) and normalized variability (NV).

Techniques are disclosed for identifying anomaly object types during classification of foreground objects extracted from image data. A self-organizing map and adaptive resonance theory (SOM-ART) network is used to discover object type clusters and classify objects depicted in the image data based on pixel-level micro-features that are extracted from the image data. Importantly, the discovery of the object type clusters is unsupervised, i.e., performed independent of any training data that defines particular objects, allowing a behavior-recognition system to forgo a training phase and for object classification to proceed without being constrained by specific object definitions. The SOM-ART network is adaptive and able to learn while discovering the object type clusters and classifying objects and identifying anomaly object types.

Described is a system for object and behavior recognition which utilizes a collection of modules which, when integrated, can automatically recognize, learn, and adapt to simple and complex visual behaviors. An object recognition module utilizes a cooperative swarm algorithm to classify an object in a domain. A graph-based object representation module is configured to use a graphical model to represent a spatial organization of the object within the domain. Additionally, a reasoning and recognition engine module consists of two sub-modules: a knowledge sub-module and a behavior recognition sub-module. The knowledge sub-module utilizes a Bayesian network, while the behavior recognition sub-module consists of layers of adaptive resonance theory clustering networks and a layer of a sustained temporal order recurrent temporal order network. The described invention has applications in video forensics, data mining, and intelligent video archiving.

The invention relates to a wind turbine drive chain vibration noise suppression and incipient fault feature extraction method. The method includes utilizing the experience mode to analyze single measuring point vibration signals, screening intrinsic mode functions capable of characterizing the fault feature frequency, reconstructing and acquiring fault feature signals to highlight the fault features, removing the noise effect on the constructed signals through autocorrelation analysis, combining with the wavelet packet transform feature frequency extraction method, and implementing the noise suppression of the single measuring point vibration signals; on the basis, utilizing the adaptive resonance theory to combine and analyze a plurality of measuring point vibration signal spectrum after the noise suppression treatment, and implementing the extraction of incipient fault features. The method has the advantages that the effects of background white noise and short interference noise can be removed effectively, the weak incipient fault feature frequency can be extracted, and the method can be directly applied to a wind turbine monitoring and fault diagnosis system and implement the diagnosis of wind turbine drive chain mechanical faults.

The invention discloses a neural network TCM syndrome diagnosis system based on self-adaptive resonance theory, comprising: a TCM four-diagnosis information preprocessing module, a syndrome differentiation module, a rule storage module and a visualization module. The four modules interact and support each other. The information preprocessing module of the four diagnoses of traditional Chinese medicine is connected with the syndrome judgment module, and the input vector of the processing system is used to guide the dynamic establishment of the syndrome judgment module to obtain the syndrome differentiation rules; the rule storage module is connected with the information of the four diagnosis of traditional Chinese medicine. The preprocessing module and the syndrome differentiation module establish a two-way connection relationship, providing empirical rules for the two, and the latter two read and modify the empirical rules; the visualization module is connected with the information preprocessing module Visualize diagnostic information and diagnostic rules. The invention can quickly perform incremental matching learning on new case samples, and apply the improved dialectical model SWART2 to improve the correct rate and adaptability of the system, and introduce visualization tools to improve the humanization and interactivity of the system.

Techniques are disclosed for discovering object type clusters using pixel-level micro-features extracted from image data. A self-organizing map and adaptive resonance theory (SOM-ART) network is used to classify objects depicted in the image data based on the pixel-level micro-features. Importantly, the discovery of the object type clusters is unsupervised, i.e., performed independent of any training data that defines particular objects, allowing a behavior-recognition system to forgo a training phase and for object classification to proceed without being constrained by specific object definitions. The SOM-ART network is adaptive and able to learn while discovering the object type clusters and classifying objects.

The invention discloses a fused face recognition method based on integrated learning, which comprises the following steps: 1.) inputting a face image to be identified; 2.) identification based on ART2 (Adaptive Resonance Theory 2) face recognition method: if the ART2 network system has a returned recognition result, receiving the recognition result, which means the identification is successful, else, entering the next step; 3.) identification based on feature face recognition method: assuming that the threshold of the method is S and the identification score of the face image by the method iss, if s>S, receiving the recognition result, which means that the identification is successful, else, entering the subsequent fused identification step; and 4.) fused identification: sequencing by using the degrees of similarity which are respectively given out by the two single recognition methods, and comparing the recognition results, wherein the first ones in the sequence are identical, receiving the recognition result, which means that the identification is successful, else, the identification fails. By adopting the concept of integrated learning, the ART2 face recognition method and feature face recognition method are fused and complemented, thereby overcoming the limitations in the single face recognition method and enhancing the integral recognition performance.

Techniques are disclosed for analyzing a scene depicted in an input stream of video frames captured by a video camera. In one embodiment, e.g., a machine learning engine may include statistical engines for generating topological feature maps based on observations and a detection module for detecting feature anomalies. The statistical engines may include adaptive resonance theory (ART) networks which cluster observed position-feature characteristics. The statistical engines may further reinforce, decay, merge, and remove clusters. The detection module may calculate a rareness value relative to recurring observations and data in the ART networks. Further, the sensitivity of detection may be adjusted according to the relative importance of recently observed anomalies.

A sequence layer in a machine-learning engine configured to learn from the observations of a computer vision engine. In one embodiment, the machine-learning engine uses the voting experts to segment adaptive resonance theory (ART) network label sequences for different objects observed in a scene. The sequence layer may be configured to observe the ART label sequences and incrementally build, update, and trim, and reorganize an ngram trie for those label sequences. The sequence layer computes the entropies for the nodes in the ngram trie and determines a sliding window length and vote count parameters. Once determined, the sequence layer may segment newly observed sequences to estimate the primitive events observed in the scene as well as issue alerts for inter-sequence and intra-sequence anomalies.

The invention discloses a multi-objective multi-modal particle swarm optimization method based on Bayesian adaptive resonance. The method comprises the following steps: dividing all particles into a plurality of populations by using a Bayesian adaptive resonance theory; sorting the particles of each population according to a non-dominated sorting method and the special congestion distance; updating the particles in the population by using the individual optimization of the particles and the global optimization of the population; connecting the non-dominated solution sets of various groups endto end to form a closed loop topology, and performing local exploration by using a particle swarm optimization algorithm based on the loop topology; and repeating the two updating and exploring processes until a termination condition is met, and outputting all the non-dominated solution sets and the Pareto frontier. The method is suitable for optimization of solving the multi-target multi-modal problem, the distribution of the Pareto leading edge can be found in the target space, the corresponding Pareto optimal solution set can be found in the decision variable space, a redundant backup method is provided, and the reliability of engineering practice activities is improved.

A sequence layer in a machine-learning engine configured to learn from the observations of a computer vision engine. In one embodiment, the machine-learning engine uses the voting experts to segment adaptive resonance theory (ART) network label sequences for different objects observed in a scene. The sequence layer may be configured to observe the ART label sequences and incrementally build, update, and trim, and reorganize an ngram trie for those label sequences. The sequence layer computes the entropies for the nodes in the ngram trie and determines a sliding window length and vote count parameters. Once determined, the sequence layer may segment newly observed sequences to estimate the primitive events observed in the scene as well as issue alerts for inter-sequence and intra-sequence anomalies.

An electronic circuit system implementing and executing machine learning inference engines. While ML inference engines are based on (architectures and parameters defined by) configured, trained and tuned machine learning models, our design has the novel ability to support data driven, on-the-fly-reconfigured model runs. Reconfiguration and tuning operations include dynamic computational graph modifications, define-by-run alterations, changes to network depth (number of layers) and width (neurons per layer), and adjustments to weights, biases, plus activation function parameters. Neural networks supported include Feed-Forward, RNN, CNN, and Hopfield architectures, plus Ensemble, Federated, Cooperating, Adversarial, and Swarm collections. Decision Trees and Forests are also supported, as are more esoteric approaches such as ART and KAN. Our invention is capable of running both standalone and cooperatively, the cooperative processing being local and / or remote / cloud based, interfacing with telemetry applications to feed data, and machine learning software to feed new or updated models.

The invention is a method for detecting out-of-specification quality of a product and estimating actually measured value of the product. The method for detecting the out-of-specification quality of the product is to construct a fault detection (FD) model representing the relationship between technological parameters and the out-of-specification of the product by applying a classification and regression tree (CART) method so as to detect the condition that the process data parameters are all in the specification (normal) but the product quality goes beyond the specification on a production line in real time. The method for estimating the actually measured value of the product is to establish a measured data quality index (DQIy) model according to an adaptive resonance theory 2 (ART2) and normalized variability (NV).

Techniques are disclosed for a video surveillance system to learn to recognize complex behaviors by analyzing pixel data using alternating layers of clustering and sequencing. A combination of a self organizing map (SOM) and an adaptive resonance theory (ART) network may be used to identify a variety of different anomalous inputs at each cluster layer. As progressively higher layers of the cortex model component represent progressively higher levels of abstraction, anomalies occurring in the higher levels of the cortex model represent observations of behavioral anomalies corresponding to progressively complex patterns of behavior.

The invention discloses a cell interruption detection positioning method based on an adaptive resonance theory. The method comprises the following steps of: in an acquisition region including a plurality of cells in, acquiring kPI, user identification and position information of a user in a mode that the user reports a measurement report periodically, and clustering the user by adopting a method based on an adaptive resonance theory (ART) network and a K-S test (Kolmogorov-Smirnov test) according to the KPI information; ; and judging and positioning an interrupted cell in an acquisition area by combining user position information according to a clustering analysis result. According to the invention, based on the ART method, the condition that there are few or even no tagged historical datain the network is fully considered, and can can accurately and effectively complete the cell interruption detection function in combination with K-S test method, and the method has a good positioningeffect on the interrupted cell by combining the position information.

Techniques are disclosed for detecting the occurrence of unusual events in a sequence of video frames Importantly, what is determined as unusual need not be defined in advance, but can be determined over time by observing a stream of primitive events and a stream of context events. A mapper component may be configured to parse the event streams and supply input data sets to multiple adaptive resonance theory (ART) networks. Each individual ART network may generate clusters from the set of inputs data supplied to that ART network. Each cluster represents an observed statistical distribution of a particular thing or event being observed that ART network.

A sequence layer in a machine-learning engine configured to learn from the observations of a computer vision engine. In one embodiment, the machine-learning engine uses the voting experts to segment adaptive resonance theory (ART) network label sequences for different objects observed in a scene. The sequence layer may be configured to observe the ART label sequences and incrementally build, update, and trim, and reorganize an ngram trie for those label sequences. The sequence layer computes the entropies for the nodes in the ngram trie and determines a sliding window length and vote count parameters. Once determined, the sequence layer may segment newly observed sequences to estimate the primitive events observed in the scene as well as issue alerts for inter-sequence and intra-sequence anomalies.

The invention discloses a predictive maintenance method for an uninterruptible power supply (UPS). The method comprises the following steps: collecting signal characteristics of a preset type of UPS equipment in a machine room; performing dimension reduction processing on the preset type of signal features by using a principal component analysis method to obtain main signal features; predicting the real-time state of the UPS equipment by adopting a simplified fuzzy adaptive resonance theoretical graph neural network according to the main signal characteristics; predicting the residual servicelife of the UPS equipment by adopting a multi-scale convolutional neural network according to the main signal characteristics; and matching a corresponding maintenance decision for the UPS equipment according to the real-time state and the prediction result of the residual service life. The invention further discloses an electronic device and a computer readable storage medium. Therefore, maintenance decision suggestions of the single UPS equipment and the whole machine room can be given according to various indexes designed under different conditions, accidents or excessive maintenance of theUPS equipment is prevented, the maintenance strategy of the whole machine room is adjusted in time, and the maintenance effect is improved.

The invention discloses a fused face recognition method based on integrated learning, which comprises the following steps: 1.) inputting a face image to be identified; 2.) identification based on ART2 (Adaptive Resonance Theory 2) face recognition method: if the ART2 network system has a returned recognition result, receiving the recognition result, which means the identification is successful, else, entering the next step; 3.) identification based on feature face recognition method: assuming that the threshold of the method is S and the identification score of the face image by the method iss, if s>S, receiving the recognition result, which means that the identification is successful, else, entering the subsequent fused identification step; and 4.) fused identification: sequencing by using the degrees of similarity which are respectively given out by the two single recognition methods, and comparing the recognition results, wherein the first ones in the sequence are identical, receiving the recognition result, which means that the identification is successful, else, the identification fails. By adopting the concept of integrated learning, the ART2 face recognition method and feature face recognition method are fused and complemented, thereby overcoming the limitations in the single face recognition method and enhancing the integral recognition performance.

The invention provides a breast milk feeding amount intelligent calculation and dynamic adjustment method and system, and relates to the technical field of intelligent feeding, and the method comprises the steps: collecting real-time physiological parameters of an infant, constructing an infant digital model through a multi-scale morphological neural network, calculating a basic energy consumption value through combining with a biological rhythm prediction model, and calculating the feeding amount of the infant. The standard breast milk feeding amount is calculated through the self-adaptive fuzzy neural network; the method comprises the following steps: collecting mother physiological state data and breast milk component data, calculating a breast milk nutrition density coefficient by using a deep mixing cognitive network and an immune evolution self-organizing network, correcting a breast milk reference feeding amount, and finally calculating an actual breast milk feeding amount through a self-adaptive resonance theory network; growth and development data of an infant after breast milk intake are collected, a multi-layer spiral pulse neural network is used for extracting development feature vectors, a development deviation value is calculated based on a fractal dynamics model, network parameters are dynamically adjusted through a biological group intelligent optimization system, and dynamic optimization of breast milk feeding amount is achieved.

The invention relates to the technical field of vibration monitoring, in particular to a rotating machine vibration protection method and system based on particle swarm optimization, and the method comprises the steps: obtaining a corrected load torque based on a polynomial fitting temperature compensation algorithm in combination with a real-time environment temperature and a sensor output voltage, and synchronously calculating an actually measured vibration amplitude; configuring the number of variational mode decomposition layers according to the real-time rotating speed of the rotor to obtain a preprocessed vibration signal, adjusting the number of independent component analysis iterations according to the load torque, separating the preprocessed vibration signal to obtain independent vibration components, and generating a feature vector; calibrating adaptive resonance theory network parameters based on a particle swarm optimization algorithm, obtaining a normalized feature vector, calculating the similarity with a fault feature template, outputting a diagnosis result, calculating the deviation between an actually measured vibration amplitude and a normal vibration threshold, and performing compensation according to the diagnosis result and the deviation. According to the scheme, the fault diagnosis accuracy of the rotary mechanical vibration protection system is effectively improved through multi-working-condition self-adaptive diagnosis.

The invention discloses a fault diagnosis system for an aviation turboshaft engine, and relates to the technical field of fault diagnosis. An acquisition module acquires horizontal, vertical and axial vibration signals; the processing module is used for generating real-time feature vectors by introducing dynamic short-time Fourier transform, variational mode decomposition and singular value decomposition of an unfixed-width Gaussian window; the incremental diagnosis module calculates the matching degree between a real-time feature vector or a prediction feature vector and a known class through a dynamic expansion network, selects to belong to the known class or triggers an expansion mechanism according to an adaptive resonance theory, constructs joint loss by using a Bayesian learning framework after expansion, and performs Gaussian process modeling; loss is determined through Bayesian optimization, and network parameters are updated; the prediction module generates a prediction feature vector when there is no fault; the maintenance module sends or stores a maintenance scheme according to needs, self-adaptive recognition of known classes and new fault classes is achieved, the state can be pre-judged in advance, the operation safety of the engine is effectively guaranteed, and the maintenance cost is reduced.